083 - Understanding Knock

I was asked at the end of this particular webinar about the use of the 'Hamming' or 'Blackman' algorithm for calculating knock window by BarryG, and I wanted to add some more to this discussion here. I haven't used these algorithms and hence couldn't answer Barry's question during the webinar. I also wanted to clarify this point as I discussed the knock window in regard to the position in the engine cycle where knock is likely to occur in response to this question which in hindsight was a little off tangent so let me elaborate.

I believe what Barry was getting at was a way of finding the knock frequency for a specific engine combination which is a valid question and one that I didn't cover in the webinar. I've always held the belief that we don't need to be mathematicians in order to tune engines, and I still feel this is accurate. It doesn't hurt if you know your way around a calculator though.

It's important to know what the knock frequency of your engine is so that you can correctly adjust the frequency filtering of the knock detection system and there are two techniques I use to find this frequency. The first and quickest/simplest is to calculate the frequency based on the bore diameter. You can do this using the equation below:

Knock frequency (kHz) = 1800 / (3.14 x Piston Dia (mm))

This will get you a pretty good starting point in most cases and I've personally found it to work well. When using an ECU like the Motec M1, we can define and monitor 4 knock frequencies simultaneously and this allows us to test multiple frequencies around the calculated knock frequency and find which gives the clearest knock signal. It's a little trial and error but is very easy to do and was demonstrated in this webinar: https://www.hpacademy.com/previous-webinars/motec-m1-knock-control-example/

I'll point out that many current generation factory knock sensors are narrow band sensors that are tuned to be most sensitive around the second order knock frequency to improve the signal to noise ratio. This means that we want to set the knock frequency to double the calculated frequency.

The other option is to actually measure the knock frequency for your particular engine so you know exactly where knock occurs. If you're using an audio knock detection unit you can take the output from this and connect it to the mic input on your laptop. You can then use a piece of audio editing software such as GoldWave to record an audio sample while the engine is encountering knock. The audio sample on it's own isn't that useful and what we need to find is the frequency that the knock is occurring at. To find the knock frequency we need to transfer between the time domain and the frequency domain and we do this using the Fast Fourier Transform (FFT) function in the audio editor. This will now display the audio sample in terms of the frequency range and the intensity at various frequencies is shown via colour. This makes it quite easy to find the specific frequency that knock is centred at.

If there is enough interest I'll look at performing a webinar showing how to do this.

I missed the live version of this webinar and had two questions. I figured since the title of this thread is specific to this webinar I'd ask my questions here.

1. How successful is water/meth at suppressing knock. I've seemed to have decent luck with it on street cars but curious of your opinion.

2. Are you able to force the engine to knock safely on a dyno without load? If so what should be the method?

1. Water/meth injection is pretty well proven to help reduce combustion temperatures and therefore reduce the chance of knock. How effective it is will depend on a number of factors though and from our own testing it is probably more effective on high boost applications. Check out a webinar we did on it here: https://www.hpacademy.com/previous-webinars/069-watermeth-injection-tuning-motec-m150/

2. No, you're going to need some load on the engine to produce knock. Again there is a webinar setting up knock control on the Motec M1 that might be useful to demonstrate this: https://www.hpacademy.com/previous-webinars/motec-m1-knock-control-example/

Hi Andre. Here's a vote for a webinar on how to calculate the knock frequency more accurately.

If one is using a PLex Knock Monitor would you suggest trying the second order frequency instead of the frequency calculated by the knock monitor using the bore size or is that normally not necessary? Is there any downside to using the second order frequency?

Going a bit off topic with another question specific to the Plex. Should you set the frequency range for the audio output to be in the range of the calculated knock frequency? The high pass 10khz setting is most effective at removing engine noises but I don't know if it is also filtering out the knock as well.

General baseline = Knock Frequency = 900/(pi * 0.5 * bore in mm)

The result is in kilohertz so for example the result for a 86 bore would be 900/(3,14*0.5*86)=6.66Khz. The second order would then be "times 2" = 13,2Khz.

If you look at the webinar that's "Motec M1 setting up knock control" - there's an excellent explanation how to narrow it down to an exact frequency and use that frequency for knock control - I'm not sure if other ECUs offer the same functionality but on the M1 you can log 4 different frequencies (you still use only 1 as a knock control parameter) but you can log multiple and narrow down the frequency.

The calculation I listed for knock frequency is actually identical to yours toxin, however I use 1800 instead of 900/0.5 in the equation just for simplicity.

I haven't found any other ECUs that will simultaneously log multiple knock frequencies unfortunately as the M1 makes it very easy to dial in the exact frequency really quickly.

Hi Toxin,

Just to clarify, the M1 can use all four knock frequencies that are set for monitoring knock, this is setup using the Knock Mode Enable function. This allows the choice of using 1,2,3,or 4 of the frequencies to generate the Ignition Cylinder N Knock Level.

If I recall correctly BlackRex, the M1 knock control webinar was on the old package where you could log/monitor 4 frequencies but only frequency A was used for knock detection. As you say, all 4 frequencies can now be utilised.

If you did it on the 86, then yes, you are only using channel A for the actual knock control signal, and the other three are for monitoring.

Great webinar Andre, thanks.

I wasn't able to tune in live, a question I have is regarding a safe margin with timing when tuning without knock control

How do you decide on an appropriate amount of timing to pull to provide a safety margin?

e.g. on 98 pump fuel, say you encounter the knock limit at 3.5k with 180kpa boost and say 25deg. How much timing would be sensible to pull out to create a safe margin?

Another question, on an high boost engine that is knock limited across the rev range, how do you approach safely optimising ignition tuning at higher rpm and load points, where damage can occur really quickly?

Thanks

Alex

@Error404 the margin I tend to leave will depend to a degree on how much power the engine is producing and how it will be used. The first step is to actually remove timing until the knock is eliminated - Essentially finding where the knock threshold is. Beyond this I'd generally suggest retarding the timing by a margin of 2-3 degrees to build in a safety barrier. As I mentioned in the webinar, it's worth doing a few pulls with a much longer ramp time (perhaps 14-18 seconds) once the tune is complete to load the engine a little harder and build more heat in the combustion chamber. This will better replicate the conditions on a race track for example and if you're still knock free under these operating conditions you can be confident your timing is safe.

As for your second question, it really depends on the specific power level. For example on a drag engine producing maybe 1000 whp from 2.0 litres, almost any level of knock could be damaging. In this case a little caution is required and I tend to be a little conservative on the ignition advance, preferring to use more boost and limit the timing to where I start to see the torque increase per degree of timing start to plateau as I continue adding timing.

On a less powerful engine, you're unlikely to have the engine instantly fail at the first hint of knock (although obviously this depends on the strength of the engine components in the first place). The approach I use is the same as above - Use a little more boost and a little less timing and be conservative with your ignition advance.

Regardless of the above, my aim is always to find the knock threshold at lower boost/power levels initially where it's less likely to cause damage. This will guide you as to the timing the engine wants - As you increase boost the timing will generally retain the same shape through the rev range but understandably we will be retarding the timing as boost increases.

Before increasing the boost I would start by tuning ahead in the ignition table, removing perhaps 2-3 degrees per 20 kPa as an approximation. As I increase the boost pressure I'm listening for knock and I'm ready to instantly abort the run if I hear anything. In the event that the engine does knock, I'd abort the run, further retard the timing, and then try again.

Thanks for the very detailed reply!

Hi Toxin and Andre,

My question wasn't about how to calculate the knock frequency but whether you generally should use the 1st order or the 2nd order frequency. Is there a downside to using the second order, such as maybe missing out some true knock events?

Most current OE knock sensors seem to be narrow band sensors that are tuned to the second order frequency as a rule. The upside is that this improves the signal to noise ratio and knock is typically more clearly defined. The downside is that the amplitude is lower at the second order which can make it more difficult to detect.

The best solution is to simply test at the calculated (or measured) knock frequency, as well as the second order frequency and see which shows a better distinction between knock and background noise.

Or you can use an M1 and a wide band sensor, and allocate two channels to the first order, and two to the second order frequencies, and have it covered.

Hi.

I have the plex knock monitor and it works great.

I have had some issues lately, especially i had two 2JZ stock head engines that made extremely much noise at load/boost. I tried different frequency filters but wasnt able to overcome the issue.

The noise curve at the plex was pretty flat and linear from down load to midrange but then increase at about 80dg angle to a noise value of almost 5x the level down low.

The same noise made it imposible for me to use my hearing to identify knock.

To comparison i recently did another built 2JZ with solid bucket valve shims and that one was perfect noise spectrum on, absolutely linear noise curve and the smallest knock was crystal clear both to my ears and monitor trend.

Any good ideas on this kind of situation?

how can i be confident there is no knock when there is so much noise to my ears, can i trust that actual knock would still "stick out" on the monitors trend log?

This can prove to be a problem on some engines, although if you're dealing with properly built engines it certainly should be rare. The advantage of the knock monitor or electronic knock detection is that you are able to focus on the specific frequency that knock occurs at which 'should' reduce the effect of background noise. Of course this isn't infallible as some of the mechanical noise you're hearing may also occur at the same frequency as knock.

If I'm battling a noisy engine, I'll start by making a dramatic reduction in timing at the area I'm concerned with and comparing the results to a run with normal timing. If you're just dealing with engine noise, you'll see no change in the noise profile. If however the engine was suffering from knock then you should see a reduction in the noise profile. The other option which is understandably a little more dangerous is to purposely advance the timing to generate knock and compare that to your previous noise profile.

Quote: "If I'm battling a noisy engine, I'll start by making a dramatic reduction in timing at the area I'm concerned with and comparing the results to a run with normal timing. If you're just dealing with engine noise, you'll see no change in the noise profile. If however the engine was suffering from knock then you should see a reduction in the noise profile. The other option which is understandably a little more dangerous is to purposely advance the timing to generate knock and compare that to your previous noise profile."

That is exactly how ive come to deal with this also so thats reasuring. Other than that id say that the more time i get to use and experiment with the PLEX under different scenarios/engines the easier it is to find the best settings to get the best possible filtering and monitoring.

Ive also connected the PLEX to the dyno so that i get the knock counts vs rpm trended there as well as alarm and run-abort functions in use when i feel i have it set up right and move to full power tuning.

Hi Sir, I had one question regarding tuning for Dodge and Ford sometimes I saw knock reading on minus (sample -2) what does it means?

@AVEPOWER if you're seeing this in a log I'll assume you're seeing the ECU retard timing in response to a knock event. I've had nothing to do with tuning on the Dodge platform though so I can't give you much more information. As always I'd recommend confirming what you're seeing in the log file with audio knock detection equipment.

Hey Andre,

Have you ever come across anybody using, or used yourself, "ion" based knock detection like is found in Saab cars? Having a voltage correlated to in cylinder pressures seems like a pretty handy tool. I've done a fair amount of googling and haven't heard much talk about them for aftermarket use which surprised me. Do you have any opinions on it?

-Derek

I have a rather specific question, i have read it is quite difficult to accurately monitor knock for type 1 air cooled vw engines just due to noise. I am currently building a 2110cc turbo super beetle for a friend and trying to decide if it is worth the money or to just tune it very conservative on timing. This will be my first full ground up tune. Hopefully someone with experience on these engines can have some info. Thanks!

-Raymond

@Stingy - Apologies I missed your question. I'm aware of the ion based knock detection technique but I've had no experience with it myself and as far as I'm aware nobody in the aftermarket is using it.

@Motoray - I haven't had any experience with air cooled VW engines sorry. Given the distinction between engine noise and the knock frequency I'd expect that monitoring knock would still be very effective.

The Plex knock monitor is a impressive device and I can see it's a must have for a professional tuner. However its price is a bit scary for beginners. The initial buget of HP tuners + wide band is more than doubled after adding this sensor.

So here comes my question. Based on your guys experiences, how does the GM knock detection strategies (E38) correlates to the Plex monitor ?

Would my beloved LS3 engine be safe if I tune it after a OTR cold air intake + headers just trusting the ECU knock detection?

Regards,

I've tuned plenty of GM LS V8 engines over my career and I'm happy to report that the knock control in the factory PCM is usually very good. I'd generally rely on it for the majority of my GM reflash tuning unless we were doing something that was pushing the boundaries in terms of power. I will say that I've had the odd occasion where the knock control was overly sensitive but this was rare. You can test to see if the knock the scanner is picking up is real by retarding the timing and seeing if that removes the knock, or richening the AFR in the area you're seeing knock.